Catalysing the Primordial Soup of Multi-Agent DeFi
An Incentive Model to Kickstart Fetch Agent DeFi Development at Scale
This post is aimed at people who grok the potential of machine agents in DeFi and who are looking to help bootstrap it into existence. If you’re not familiar but interested to learn more, reach out.
Superior Emergent Financial Services
It's my belief that in Fetch.Ai's groundbreaking agent framework the ingredients required to make a radically more efficient economy are in place. What's not there yet is the trigger to catalyse a swarm of data scientists to build out the necessary components and enable the novel emergent services I describe above.
The problem? Data science is one of the best-paid and most in-demand sectors in the world. It’s difficult to get data scientist time. But I have hope. Why? Because crypto has a strong track record, and arguably its killer app, is in incentivising expensive supply sides. We've seen this with Bitcoin and Ethereum miners, we've seen it with liquidity mining and the 10s of billions now locked in DeFi and most relatedly we've seen a crypto project in Numerai build a community of data scientists by giving away $NMR in exchange for market models and signals.
Now, Fetch has obviously been making headway with its incentivised testnets, and the number of people running agents is mounting well. However, I suppose I'm keen to see an acceleration of adoption of agents in DeFi specifically, and I think this requires a concerted and deliberate effort. So in this post I describe some context to the problem and some economic design ideas to help bring #FetchFi to the fore.
The Infrastructure of Multi-Agent DeFi
My hypothesis is that FetchFi requires a critical mass of an underlying ecosystem to be built in order to start creating *really* novel emergent financial services.
What constitutes that ecosystem? Partly it's infrastructure, partly it's education for DeFi builders. I am focused on the infrastructure.
The infrastructure revolves around the components of the agent framework. That includes skills – behaviours and handlers – as well as protocols and agents themselves.
Sidebar: Some components are data services which should not be in the direct remit of the Fetch ecosystem. For example, an API to predict the risk/reward of Uniswap pools. Perhaps this kind of component should have its own incentive bootstrapping model. Now that I think of it, this is probably what Ocean's data services should be 🤔. Why are they not emerging on Ocean yet?
Myself and other folks are working on some of this stuff behind the scenes. If we create a successful agent and show it to the world, this will likely accelerate uptake, but could we accelerate things even further, and even extend Fetch's tokenomics at the same time?
Marketplace for #FetchFi Components
I propose a marketplace for #FetchFi components. In this marketplace, agents, skills and other components are tokenised.
Developers are charged a fee to use agents and skills. Thus, components are cashflow-producing.
In the early days, component developers will have weak sales. This will not be attractive – it will not justify the cost of their time. To motivate them, emit a token based on the performance of their components – more usage, more tokens.
This has the benefit of building out the emerging base layer of infrastructure required for valuable, novel emergent services to become established on top. It overcomes a chicken-and-egg situation where it's initially difficult to build agents which compete with standard DeFi services.
Over time, knowledge about which components are most critical will be fed back into market prices by builders. This creates a positive feedback loop where high-quality and relevant components get surfaced to the top.
The implementation details of the marketplace could be left open – I'd point interested readers to the design and success of Ocean Market built on Balancer as a potential source of inspiration. All I would suggest as critical in the design is that it uses the network's native token as the settlement asset – i.e. that the component token is pooled alongside this protocol's token.
This leads nicely into the token itself. Although I think this could be done with $FET itself, it would make sense for the network to have its own token. There's value in creating a separate token because it creates ownership and focus within the Fetch community around the goal of extending agents into DeFi.
The new token could be merged with FET at a later date. How would this work? Whenever fees are taken on the marketplace, a portion is fed into the treasury. Over time, the tokens in the treasury are converted into FET, and TKN is burned proportionally. With time, the treasury is entirely in FET. TKN holders can then vote to transition to using FET as the settlement asset on the marketplace and to use FET as the governance token, thus completing a merge.